An appeals court has ordered a man convicted of manslaughter to be resentenced, after finding that the judge in his original sentencing hearing was affected by an AI-generated video of the victim.

The case centers on a digitally reconstructed video of the deceased, produced with artificial intelligence, that was presented during sentencing proceedings. The appellate court determined that the AI-generated video influenced the original judge's decision, undermining the validity of the sentence handed down.

The ruling makes explicit what has until now been largely a theoretical concern in legal and media circles: synthetic video of a real person — in this case, a crime victim — can carry enough emotional and evidentiary weight to alter a judicial outcome. For an industry already grappling with the licensing, provenance, and authenticity questions raised by generative AI, this is one of the first documented instances of an AI-generated likeness directly affecting a consequential real-world decision by a court.

The defendant, previously convicted of manslaughter, will now face a new sentencing hearing. The appeals court's finding stops short of invalidating the conviction itself; the issue is narrowly confined to how the AI video shaped the sentencing judge's reasoning.

For working photographers and media businesses, the decision lands amid ongoing debate over content credentials, C2PA-style provenance standards, and the admissibility of synthetic imagery. Courts have historically treated photographic and video evidence as reliable ground truth. A ruling that treats an AI-generated video as a factor capable of distorting a sentencing outcome signals that this assumption no longer holds unchallenged.

The case also raises unresolved questions about who produces such videos, under what authorization, and with what disclosure. Unlike courtroom sketch artists or victim-impact statements delivered by family members, an AI-generated video of a deceased person cannot be cross-examined and rests entirely on the choices made by whoever generated it — the script, the likeness, the voice, the emotional register.

Legal observers expect the resentencing to proceed without the AI video in evidence, which will offer a rough real-world test of how much the synthetic material actually moved the original outcome. Further appellate decisions on AI-generated victim content are likely to follow as courts confront similar requests from grieving families and prosecutors.

The resentencing has not yet been scheduled, and it remains to be seen how courts will draw the line between memorial tributes and evidence as generative video tools become cheaper and more widespread.